behaviorial finance

IT IS

ILLE

GAL

TO R

EPRODUCE T

HIS A

RTICLE

IN A

NY FO

RMAT

16 MODERN PORTFOLIO THEORY AND BEHAVIORAL FINANCE FALL 2004

Traditional finance assumes that we arerational, while behavioral finance simplyassumes we are normal.

—Meir Statman1

There is nothing quite like a NobelPrize to focus the investing public’sattention. Harry Markowitz devel-oped the concept of mean variance

optimization in the early 1950s,2 but it wasn’tuntil he was awarded the Nobel Prize in Eco-nomics in 1990 that most investors began todevelop a keen interest in efficient frontiers. In2002, Daniel Kahneman won the Nobel forhis work in behavioral finance, and suddenlyinvestors everywhere are looking at behavioralfinance techniques to improve their risk-adjusted performance. In fact, Kahneman (andhis colleague, Amos Tversky)3 did their keywork back in the 1970s.4

Well, better late than never. The fact isthat Markowitz’ work revolutionized the waythe capital markets operated and what theimplications of that knowledge might be forthe design and diversification of investmentportfolios. Before mean variance optimizationcame along, our idea of diversification was tohold five stocks when one might have seemedto suffice. Markowitz pointed out that it isn’tsimply the number of different securities weown that matters—it is the correlation of thosesecurities with each other that matters. Thus,if we own BP Amoco and Royal Dutch Shell,we are far, far less diversified than if we own

BP Amoco and Microsoft, simply because thefactors that affect those two companies are,while not completely distinct, actually quitedifferent. (In modern portfolio theory terms,the covariance of BP Amoco and Microsoft islower than the covariance of BP Amoco andRoyal Dutch Shell.)

And while it is too soon to say with cer-tainty, it seems likely that the work of thebehavioral finance professionals will similarlyrevolutionize the way we think about thedesign and management of our portfolios.Since Markowitz won the Nobel in 1990, wehave tended to design our portfolios as thoughwe were all mean variance optimizers, perfectlittle economic beings who always made the“right” decision in our own interests. Thework of Kahneman, Tversky, et al. has blownthis cozy little conceit to pieces. True, some-times we behave like perfect economic beings.But other times we behave like, well, humanbeings. We make decisions on the basis ofbiases that don’t reflect real world facts. Weallow our responses to decisions to depend onhow the questions are framed. We engage incomplex mental accounting, ignoring the factthat our various asset baskets are all interre-lated. We allow ourselves to be driven by hopesand fears, rather than facts.

So which is better—modern portfoliotheory, which describes how markets work,or behavioral finance, which describes howpeople work? The answer, of course, is thatwe need both. MPT and behavioral finance

Modern Portfolio Theory and Behavioral FinanceGREGORY CURTIS

GREGORY CURTIS

is chairman of Greycourt &Co., Inc. in Pittsburgh, [email protected]

It is illegal to reproduce this article in any format. Copyright 2004.

are both important tools in helping us design and managesuccessful investment portfolios. Both have advantagesand disadvantages, however, and it is useful to reviewthose pros and cons before we proceed to think abouthow combining the two approaches might work.

LIMITATIONS OF MPT AND BEHAVIORAL FINANCE

Limitations of Modern Portfolio Theory

The heart has its reasons, of which reason knows nothing.

—Pascal

I have elsewhere5 identified the many issues pre-sented by modern portfolio theory, especially when wetry to apply its theories to the real world of investmentportfolios. Suffice it to say that MPT is a theoretical con-struct that attempts to describe how capital marketsoperate, not a recipe for designing investment portfolios:

[M]odern portfolio theory is useful under certainprescribed conditions, some of which we knowabout and some of which we don’t. We know, forexample, that MPT assumes continuous pricing, aworld in which markets are free, societies are freeand stable, and investors are rational wealth-max-imizers. Events that occur outside these conditionsare not merely events that fall several standard devi-ations outside what MPT would predict. Instead,they are events that have nothing whatever to dowith MPT, but are governed instead by very dif-ferent rules that can be understood only by refer-ence to very different theories.6

In other words, MPT is descriptive, not prescriptive.And even insofar as MPT can be said to be prescriptive,its predictive accuracy about how markets will behave inthe future is unusably low within any kind of time horizonrelevant to human investors. Finally, MPT’s assumptionthat we are all and always rational wealth-maximizers isclearly incorrect.

As a result of these issues, when financial advisorsattempt to communicate with clients about their portfo-lios using MPT constructs, communication largely ceases.As Statman hilariously puts it:

Conversations with clients often resemble the GaryLarson cartoon in which the man says to his dog,“Ginger, I have had it! Stay out of the garbage,Ginger. Understand, Ginger? Stay out of thegarbage, or else!” And Ginger hears: “Blah blahblah, Ginger. Blah blah blah, Ginger. Blah blahblah, Ginger.” Financial advisors say, “High returnscannot be guaranteed. No one can guarantee thathigh risk will bring high returns. No guarantee,you understand?” And clients hear: “Blah blah blahhigh returns. Blah blah blah no risk. Blah blah blahguaranteed!”7

Limitations of Behavioral Finance

In a sense, behavioral finance picks up where modernportfolio theory leaves off, completing the circle. Itdescribes how investors actually behave, rather than howthey should behave. It recognizes that we sometimes actin our own best economic interests, and that we some-times don’t. Assuming that modern portfolio theorylargely correctly describes the way markets operate, behav-ioral finance describes how we might best profit from thatknowledge.

The foundations of behavioral finance were estab-lished by the work of Daniel Kahneman and AmosTversky, the founders of “prospect theory.”8 Prospecttheory suggests that, in making decisions (especially, butnot exclusively, financial decisions), we tend to “irra-tionally” favor long shots, to avoid near-certain gains, tobuy insurance against losses that are quite unlikely tooccur, and to take large risks to win back large losses. Thetheory focuses especially on the fact that our attitudetoward the risks associated with obtaining gains may bequite different from our attitude toward risks associatedwith losses.9

But there are serious issues associated with behav-ioral finance, just as there are with modern portfoliotheory. Before proceeding, let’s examine some of them.

Weather patterns don’t know they are being studied.Behavioral finance shares with other forms of psycho-logical research the challenge of dealing with sentientresearch subjects. If a scientist is studying weather pat-terns, the weather patterns don’t know they are beingstudied and don’t change their behavior as a result of beinginvolved in a research project. But human research sub-jects are maddeningly more complex. The very fact ofknowing that we are involved in an experiment causes usto behave differently than we would behave outside the

FALL 2004 THE JOURNAL OF WEALTH MANAGEMENT 17


laboratory context. Maybe we are answering questionshonestly and without guile. But maybe we are trying togive the investigator the answer she wants to hear. Maybewe are annoyed at having been “strong-armed” into par-ticipating in the experiment (to pass Psychology 101, forexample). Maybe we are nervous about appearing stupidand hence inadvertently give answers that we wouldn’tgive in another context.

Human subjects don’t always play by the rules.Imagine that CNN has just reported the results of a large,longitudinal10 study of 10,000 subjects showing that peoplewho ate bran muffins every day doubled their chances ofdeveloping brain tumors. Before we eliminate bran fromour diets, we might be interested to know that since thestudy went on for 10 years, almost everyone in both groupscheated. In other words, those in Group A, who weresupposed to eat bran muffins every day, didn’t—possiblybecause they got sick to death of bran muffins. And thosein Group B, who were never supposed to eat bran muffins,did in fact eat them regularly—possibly because they likedbran muffins and possibly because they figured that sincebran muffins were the subject of the research study theymust be good for you. In other words, the two groupswere a lot more like each other in the bran muffin habitsthan the researchers assumed.

The phrase “statistically significant” doesn’t neces-sarily mean much. Referring back to the study of branmuffins, when CNN tells us that the results of the studywere “statistically significant,” we tend to hear “the resultswere valid and we should stop eating bran muffins.” Actu-ally, there is a subtle but important distinction betweenstatistical significance and, say, the likelihood that theresults of the study can be applied to us. Statistical signif-icance is a measure of the randomness of the data in astudy. If the sample itself (the subjects selected to partic-ipate in the study) is biased in some way, the statisticaltests for significance will be meaningless. Thus, it wouldbe important to know that all the subjects in the branmuffin study were middle-aged female residents of theSan Francisco Bay area who were paid to participate andwho were unlikely to move away during the 10-yearperiod of the study. In other words, “statistical signifi-cance” with respect to the bran muffin study probablymeans nothing more than that similar results would likelybe generated if the study were repeated using 10,000middle-aged female residents of the Bay Area (etc.). Therelevance of the study to the rest of us remains a majorunanswered question.

What about the other 40%? If the results show that

60% of the subjects in a behavioral finance experimentmade the “wrong” choice on a financial test, what aboutthe other 40%? As investors and financial advisors, canwe simply assume that all our clients fall among the 60%who got it wrong?

Experience and education matter. All of us make baddecisions in areas we know little about, but that doesn’tmean that we will continue to make bad decisions oncewe have learned something about the subject. I might bea lousy bridge player, making one bad decision afteranother. But that’s because I don’t know much aboutbridge. If I played more often and studied the game, Iwould likely make much better decisions. And the sameis true of many of the findings of behavioral finance theory:once investors understand that their decisions are badones, and why, they are likely to make much better deci-sions in the future.

Having something really at stake matters. It’s onething to pop off with a quick answer in a laboratory set-ting. That answer might even be the “natural” answer wewould make. But it’s another thing altogether to make adecision that affects, say, a $100 million investment port-folio. In the latter case we are far less likely to “pop off,”or shoot from the hip. We are far more likely to thinkhard about what the right answers might be and to getsome advice about them.

The experimenters’ expectations affect the outcomeof their studies. Our worst fears about research studieshave proven to be correct: researchers’ expectations of whatthey will find profoundly affect what they actually find.In one study, a group of teachers was told that the mem-bers of their class had scored in the “near-genius” rangeon a set of aptitude tests. Members of that class subse-quently got excellent grades. A second group of teacherswas told that members of their class had scored well belowaverage on aptitude tests. Members of that class subse-quently got poor grades. Needless to say, the members ofboth classes were the same. And this phenomenon of“finding what we expect to find” (or worse, finding whatwe want to find) extends even to non-human subjects. Inanother study, one group of experimenters was told thatthe strain of mice they were using had been bred to beespecially intelligent. Those mice turned out to learn theirway through mazes very quickly. Another group of exper-imenters was told that the strain of mice they were usingwas particularly dense. Those mice turned out to learntheir way through mazes very slowly. The strains of micein both groups were, of course, identical.11 Thus, whileexperimenters do their best to structure their experiments



carefully and to avoid allowing their own expectations toaffect the results, investigator bias is always an issue.

Summary

In short, employing MPT techniques when weadvise investors does tend to cause financial advisors to pro-pose optimal portfolios, but the likelihood that clientswill adopt those portfolios (or stick to them) is low. Onthe other hand, employing behavioral finance techniqueswhen we advise clients tends to result in recommendedportfolios that resonate well with the clients, but whichare not likely to be optimal in terms of the relationshipbetween risk and reward. In both cases, clients may ulti-mately be disappointed—in the first case because theyfailed to follow the advice and in the second case becausethey did follow the advice.

ITERATIVE COMBINATIONS OF BOTH THEORIES

MPT can be thought of as the “rational” approachto portfolio design. Despite its limitations, it looks unblink-ingly at the way capital markets operate and suggests howwe might optimally exploit those markets to our ownadvantage. Unfortunately, the investment strategies sug-gested by MPT are often unpalatable to investors, evenwhen they are correctly understood.

Behavioral finance can be thought of as the “ara-

tional” approach to portfolio design. An “arational”approach is not necessarily “irrational.” Indeed, the“wrong” choices we make as investors may be suboptimalfrom a purely economic perspective, but those choicesoften serve deeper emotional needs.

It is interesting to speculate about the possibility ofcombining “rational” MPT and “arational” behaviorfinance approaches into one integrated advisory process.Suppose, for example, that we were to design the client’sportfolio in the traditional manner, using MPT-basedstrategic asset allocation techniques. At the same time, wecan design the client’s portfolio using techniques informedby behavioral finance. We can then compare the tworesults in an instructive way.

Step One: Design the Traditional MPT Portfolio

Strategic asset allocation12 represents the state of theart in MPT-influenced portfolio design, although in realityit is practiced by a surprisingly small group of elite advi-sory firms.13 In the Exhibit, the column labeled “MPT”suggests a portfolio that might have been designed usingstrategic asset allocation techniques. This portfolio is basedon a forward-looking view of capital markets expecta-tions and the degree of risk required to grow the fami-ly’s asset base faster than inflation, spending, taxes, and soon. It has the advantage of being likely to succeed in purefinancial terms if the family will stick with the strategy.


BehaviorAsset Class MPT Finance Difference

US large passive 15.0% 30.0% -15.0%US small passive 5.0% 5.0% 0.0%Non-US active 5.0% 5.0% 0.0%Emerging markets equity active 5.0% 0.0% 5.0%Directional hedge 15.0% 0.0% 15.0%Absolute return hedge 15.0% 5.0% 10.0%Private equity/venture 10.0% 5.0% 5.0%Real estate passive 10.0% 5.0% 5.0%High yield debt * 0.0% *Bonds 17.0% 35.0% -18.0%Cash 3.0% 10.0% -7.0%

*Opportunistic

E X H I B I TComparing MPT and Behavioral Finance Portfolios


But that is precisely the rub. The portfolio is“uncomfortable” for the family, partly because of theinclusion of asset classes they don’t completely understand(emerging markets, directional hedge), and partly becausethe portfolio will likely incur periods of short-term under-performance that will test the family’s investment patience.For these reasons, the likelihood that the family will, infact, stick with the portfolio is small. Indeed, there is aserious possibility that the family will abandon the strategyat the very worst time. The MPT portfolio elevates risk-return optimality over investor comfort, with the resultthat the family is unlikely to meet its long-term needs.

Step 2: Design the Behavioral Finance Portfolio

Despite the 30-year history of behavioral finance,few investigators have suggested concrete ways to imple-ment the learning of that branch of economics. Tradi-tionally, in trying to get a sense of a client’s tolerance forinvestment risk, financial advisors have engaged in rule-of-thumb exercises or even more bizarre techniques. Therewas, for example, the approach we might call your-age-is-your-fate: subtract your age from 100 and that’s whatyour equity exposure should be. There were the pre-designed portfolios assigned to clients according to theirage range: if you were between 30 and 40, you got Port-folio A; between 40 and 50, Portfolio B, etc. My per-sonal favorites were the bizarre questionnaires investorswere asked to complete: “Would you rather curl up witha good book or go bungee jumping?”

Recently, however, two more promising approacheshave been suggested:

• Statman has suggested that an investment portfoliobe viewed as a pyramid, with the lowest-risk goals(and associated investments) at the broad bottomand the highest-risk goals (and associated invest-ments) at the narrow top.14

• Brunel has elaborated on Statman’s suggestion byconverting Statman’s pyramid into a more traditionalportfolio design framework: Brunel invites theinvestor to quantify the relative importance of thefour traditional investment goals: liquidity, income,capital preservation, and growth.15

In the Exhibit, the column labeled “BehaviorFinance” suggests a portfolio that might result from theuse of the Statman/Brunel approach. It is a cautious port-

folio, reflecting behavioral finance’s findings about lossaversion. Since three of the four investment goals (liquid-ity, income, capital preservation) tend to lead inevitablytoward cautious strategies, and only one goal (growth)tends to lead toward more aggressive strategies, the behav-ioral finance portfolio has the likely disadvantage ofgrowing too slowly to preserve the family’s wealth overthe years. But it has the advantage of being “comfort-able” for the family, and representing a strategy the familyis likely to stick with at least until they realize that theirasset base hasn’t kept pace with their expectations. Inother words, our behavioral finance portfolio has indulgedthe family’s inherent biases, but it may have resulted in asuboptimal portfolio that elevates comfort over invest-ment success.

Merging the Two Approaches

Given this dilemma, how can we reconcile the cap-ital markets strength of MPT portfolios with the human-centered strength of behavioral finance portfolios? Onesuggestion is to show the family both portfolios, beinghonest about the advantages and disadvantages of each.The game plan for the family would be to start with aportfolio that is closer to the behavioral finance model,but with the expectation that, over time, the family woulditeratively evolve toward the MPT model.

Thus, the family might begin with the behavioralfinance portfolio but with a five-year plan (or even longer)to move toward the MPT portfolio. The column in theExhibit labeled “Difference” shows what the family wouldhave to do to make this transition. Thus, the family willhave to decrease its exposure to “comfortable” asset classeslike U.S. large-cap, bonds, and cash, and will have toincrease its exposure to less comfortable asset classes likeemerging markets, directional (long/short) hedge, abso-lute return hedge, private equity, and real estate. In addi-tion, when the spreads between high-yield bonds andTreasury bonds reach extreme levels, the family willopportunistically gain exposure to junk bonds, movinggradually out of that asset as spreads narrow.

To make this transition palatable to the family, it isnecessary that they gain experience with less comfortableasset classes gradually. If the transition is expected to occurover a five-year period, for example, it may make senseto adjust asset class exposures at the rate of 20% per year.Thus, each year 3% of the beginning U.S. large-cap expo-sure will be sold off and invested in asset classes that needto grow.



The beauty of making the transition over time is that,while the initial goal may be to have made the completetransition in five years, there is no reason why this decisioncan’t be revisited. If the family indicates concern about aparticular asset class, investments in that category can slowdown. If the family indicates comfort with a particular assetclass, investments in that category can speed up.

Challenges Associated with Making the Transition

The challenges associated with the transition from abehavioral finance portfolio to an MPT portfolio are thesame as the challenges associated with any significant port-folio transition, namely, taxes, market timing issues, andminimum account size problems (for smaller families).Regarding taxes, it is fortunately the case that most behav-ioral finance portfolios will prove to be too cautious. There-fore, the family will tend to be moving from lower-growthassets to higher-growth assets, minimizing problems asso-ciated with having to sell low tax cost basis investments.In the Exhibit, for example, the family will need to reduceits bond and cash portfolios most significantly, and thattransition is unlikely to incur serious tax problems.

Market timing is always an issue in transitioning aninvestment portfolio. But by making the transition overan extended period of years, the family is not only get-ting used to uncomfortable assets, but is able to average-in and out of asset classes over time, minimizing the riskof bad market timing calls. More sophisticated families,of course, may wish to key the portfolio transition not to“time” but to “valuation,” i.e., moving toward neededasset classes as they become undervalued and moving outof unneeded asset classes as they become overvalued.

Minimum account size issues can also arise, ofcourse. If the family asset base is not large, it may be dif-ficult to gain a small starting position in certain asset classes.If the family has $10 million of investable assets, forexample, it should not be difficult for the family to put$1.5 million productively to work in directional hedge(via a fund of funds, for example). But if the family ismoving from a 0% directional hedge exposure to a 15%exposure over five years, it may be more problematic toput $300,000 to work productively each year. Fortunately,minimum account sizes are falling, even among some ofthe better funds of hedge funds, and in other asset classesit will be relatively easy to find an institutional mutualfund (or Delaware business trust) to invest in until thefamily meets the minimum size for a separate account.

SUMMARY

Modern portfolio theory represents the best learningwe have about how capital markets actually operate, whilebehavioral finance offers the best insights into howinvestors actually behave. But markets don’t care whatinvestors think of as risk, and hence idiosyncratic ideasabout risk and what to do about it are bound to harmour long-term investment results. On the other hand,Daniel Kahneman, Amos Tversky, and their followershave demonstrated beyond doubt that we all harboridiosyncratic ideas and that we tend to act on them, regard-less of the costs to our economic welfare.

By combining both MPT and behavioral financemodels in our work with family investors, we stand thebest chance of designing, implementing, and maintainingportfolios that will prove acceptable to our clients andthat will prove productive to our clients’ wealth.

ENDNOTES

1Statman is the Glenn Klimek Professor of Finance atSanta Clara University. Quoted in Jean Brunel, “Revisitingthe Asset Allocation Challenge Through a Behavioral FinanceLens,” The Journal of Wealth Management, Fall 2003, p. 10.

2Harry Markowitz, “Portfolio Selection,” Journal ofFinance, 1952.

3Tversky had the misfortune to die in 1996. Nobel Prizescan only be awarded to living persons.

4See, e.g., Daniel Kahneman and Amos Tversky, “On thePsychology of Prediction,” Psychological Review, 80 (1973), pp.237-25l; Amos Tversky and Daniel Kahneman, “Availability: AHeuristic for Judging Frequency and Probability,” Cognitive Psy-chology, 5 (1973), pp. 207-232; Amos Tversky and Daniel Kah-neman, “Judgment under Uncertainty: Heuristics and Biases,”Science, 185 (1974), pp. 1124-1131; Daniel Kahneman and AmosTversky, “Intuitive Prediction: Biases and Corrective Proce-dures,” Management Science, 12 (1979), pp. 313-327.

5Gregory Curtis, “Modern Portfolio Theory andQuantum Mechanics,” The Journal of Wealth Management, Fall2002, pp. 1-7.

6Ibid., p. 2.7Meir Statman, “Financial Physicians,” AIMR’s Invest-

ment Counseling for Private Clients IV (No. 4, 2002), p. 5.8The term “prospect theory” apparently derives from

one’s prospects of winning a lottery. Most people believe, how-ever, that Tversky and Kahneman adopted the term mainlybecause it was catchy and would raise the profile of their work.See, e.g., Daniel Kahneman and Amos Tversky, “ProspectTheory: An Analysis of Decisions under Risk,” Econometrica, 47(1979), pp. 313-327.



9The most extensive treatments of behavioral finance areprobably Hersh Shefrin’s Beyond Greed and Fear: UnderstandingBehavioral Finance and the Psychology of Investing (Oxford Uni-versity Press, 2002), and Richard H. Thaler’s earlier Advancesin Behavioral Finance (Russell Sage Foundation, 1993).

10A “longitudinal” study is one that continues over time,rather than being a one-off project. At one extreme, a largepopulation may be studied over decades. At the other extreme,a longitudinal study might follow a relatively small group ofsubjects for a few days or weeks.

11These studies were reported in the Wall Street Journal.See Sharon Begley, “Expectations May Alter Outcomes FarMore Than We Realize,” Wall Street Journal, November 7,2003, p. B1.

12Defined by Brunel as an approach that “should cover allthe multiple locations . . . through which a wealthy family holdstheir assets and be formulated through a multi-period processdriven by after-tax results.” Brunel, op. cit., note 1, p. 10.

13A serious drawback of strategic asset allocation, insofaras the financial industry is concerned, is that it can only beaccomplished by very senior investment professionals workingone-on-one with clients in a highly customized manner. Thisis exactly the opposite of the mass approach that maximizes theprofitability of the global financial giants.

14Meir Statman, “Behavioral Finance: Past Battles, FutureEngagements,” Financial Analysts Journal, November/December1999, pp. 18-28.

15Brunel, op. cit., note 1, pp. 11-18 passim.

To order reprints of this article, please contact Ajani Malik [email protected] or 212-224-3205.



behaviorial finance

Documents